Adding CNN model

app.py

@@ -1,7 +1,48 @@
 import gradio as gr
+import torch
+# from lr_ed.model import CNNEmotinoalClassifier
+import torchaudio
+import IPython.display as ipd
+from torch import nn
+from model import CNNEmotinoalClassifier
 
-def greet(name):
-    return "Hello " + name + "!!"
+model = CNNEmotinoalClassifier()
+model.load_state_dict(torch.load('./cnn_class_17.pt'))
+model.eval()
 
-iface = gr.Interface(fn=greet, inputs="text", outputs="text")
+to_melspec = torchaudio.transforms.MelSpectrogram(
+    sample_rate= 22050,
+    n_fft = 1024,
+    hop_length = 512,
+    n_mels=64
+)
+
+def _get_right_pad(target_waveform, waveform):
+    target_waveform = target_waveform
+    waveform_samples_number = waveform.shape[1]
+    if waveform_samples_number < target_waveform:
+        right_pad = target_waveform - waveform_samples_number
+        padding_touple = (0, right_pad)
+        waveform_padded = nn.functional.pad(waveform, padding_touple)
+    else:
+        waveform_padded = waveform
+    return waveform_padded
+
+def get_probs(audio):
+    emotions = ['happy', 'angry', 'sad', 'neutral', 'surprised', 'fear']
+    emotions = sorted(emotions)
+    
+    sr, waveform = audio
+    waveform = _get_right_pad(400384, waveform)
+    input_x = to_melspec(waveform)
+    input_x = torch.unsqueeze(input_x, dim=1)
+
+    probs = model(input_x)
+    prediction = emotions[probs.argmax(dim=1).item()]
+    return dict(zip(emotions, list(map(float, probs[0]))))
+
+mic = gr.Audio(source="microphone", type="numpy", label="Speak here...")
+label = gr.Label()
+
+iface = gr.Interface(fn=get_probs, inputs=mic, outputs=label)
 iface.launch()

cnn_class_17.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:776cb5ff06d6d2ddacef3ee8fdb61c67f241f82777c3f8a830d7c338256f174e
+size 16823379

model.py

@@ -0,0 +1,67 @@
+import torch
+import torchaudio
+import numpy as np
+from torch import nn
+from torchsummary import summary
+
+class CNNEmotinoalClassifier(nn.Module):
+    def __init__(self):
+        super(CNNEmotinoalClassifier, self).__init__()
+
+        # conv : 4, flatten, linear, softmax
+
+        self.conv1 = nn.Sequential(
+            nn.Conv2d(in_channels=1, out_channels=16, kernel_size=3, stride=1, padding=1),
+            nn.ReLU(),
+            nn.MaxPool2d(kernel_size=2)
+        )
+
+        self.conv2 = nn.Sequential(
+            nn.Conv2d(in_channels=16, out_channels=32, kernel_size=3, stride=1, padding=1),
+            nn.ReLU(),
+            nn.MaxPool2d(kernel_size=2)
+        )
+
+        self.conv3 = nn.Sequential(
+            nn.Conv2d(in_channels=32, out_channels=64, kernel_size=3, stride=1, padding=2),
+            nn.ReLU(),
+            nn.MaxPool2d(kernel_size=2)
+        )
+
+        self.conv4 = nn.Sequential(
+            nn.Conv2d(in_channels=64, out_channels=128, kernel_size=3, stride=1, padding=2),
+            nn.ReLU(),
+            nn.MaxPool2d(kernel_size=2)
+        )
+
+        self.flatten = nn.Flatten()
+
+        self.fully_connected = nn.Sequential(
+            nn.Linear(128 * 5 * 50, 128),
+            nn.ReLU(),
+            nn.Linear(128, 64),
+            nn.ReLU(),
+            nn.Linear(64, 32),
+            nn.ReLU(),
+            nn.Linear(32, 16),
+            nn.ReLU(),
+            nn.Linear(16, 6)
+        )
+
+        self.softmax = nn.Softmax(dim=1)
+
+    def forward(self, x):
+        x = self.conv1(x)
+        x = self.conv2(x)
+        x = self.conv3(x)
+        x = self.conv4(x)
+        x = self.flatten(x)
+        logits = self.fully_connected(x)
+        probs = self.softmax(logits)
+        return probs
+
+
+if __name__ == '__main__':
+    device = ('cuda' if torch.cuda.is_available() else 'cpu')
+    model = CNNEmotinoalClassifier().to(device)
+    summary(model, (1, 64, 783))